Products and systems built end-to-end — AI infrastructure, computer vision & autonomy, streaming at national scale, and industrial automation.
One Model Context Protocol server exposing ~40 tools — chat, code, image generation, memory, ops — to every AI client, backed by a self-hosted multi-GPU rig running local LLMs via ollama and vLLM for private, low-cost inference.
Air-gapped retrieval-augmented generation so sensitive documents never leave the customer boundary — built for CUI/ITAR-sensitive environments on in-house GPU infrastructure.
Perceptual video-quality scoring and bitrate–quality optimization across HLS/DASH streams at Pluto TV / Paramount scale, on GCP and AWS — automated in Go and Python.
Machine-interface and SDK integration for industrial equipment — including a machine-vision inspection proof-of-concept that helped secure a $20M+ program for a major manufacturer.
A private AI datacenter-in-a-box — RTX 5090/4090/5080 serving chat, coding, and image models over the LAN, with health monitoring and MCP tooling, so client work runs without per-token cloud costs.
Custom-trained YOLO models for real-time object detection and tracking on live video — dataset curation, training, and GPU-accelerated inference tuned for the edge.
Drones that see: onboard AI vision locking onto and tracking moving targets in flight — detection, target lock, and flight/gimbal control loops running live on the aircraft.
From images to usable geometry — AI-assisted 3D mesh generation and reconstruction pipelines producing clean, textured meshes ready for downstream tooling.